The Legibility Trap
There's a pattern I keep seeing, in tech and beyond: someone looks at a messy, complicated system and decides to make it "legible." They create dashboards, metrics, standardized processes. Everything gets cleaned up, categorized, measured.
And then, somehow, things get worse.
What Legibility Costs
James C. Scott wrote a whole book about this—Seeing Like a State—but the core insight is simple: making a system legible requires simplifying it. Simplification means discarding information. Some of that discarded information was actually important.
The classic example is forestry. German foresters in the 18th century wanted to maximize timber production. They cleared the messy natural forests and planted rows of uniform trees. Production went up—for one generation. Then the monoculture collapsed. The diversity they'd eliminated wasn't inefficiency; it was resilience.
The Metrics Trap
The tech version of this: any metric you optimize will eventually be gamed. The metric becomes the target, and the underlying thing it was supposed to measure gets ignored.
- Lines of code → developers write verbose code
- Number of deploys → teams make smaller, less meaningful changes
- Customer satisfaction scores → support teams optimize for survey responses, not actual satisfaction
Living with Illegibility
I'm not saying measurement is bad. But I am saying we should be suspicious of our urge to make everything legible. Some systems work precisely because they're messy. The mess contains information.
The alternative to legibility isn't chaos. It's local knowledge, judgment, expertise that can't be reduced to a checklist. It's trusting people to navigate complexity instead of trying to eliminate it.
Sometimes the right answer is: let it be complicated. Intervene less. Observe more.
The systems that survive long-term are rarely the cleanest. They're the ones that found ways to preserve what mattered while appearing, to outside observers, thoroughly confusing.